Public commitment surface
Editorial Calendar
What we will publish over the next four weeks, plus the next quarter as themes. When we slip, we show the slip.
♥ Subscribe via Calendar RSSThis page lists every piece Substratics has committed to publish, the section and beat each piece sits in, and a one-sentence dek on what it argues. Entries past their target date that have not shipped show a visible slip indicator with a new commitment date or a note that the piece is under review.
Three statuses are visible here: planned, published, and slipped. Drafting and in-review states are kept internal — surfacing them publicly trains a publication toward shipping marginal pieces to avoid explaining kills, which is the exact failure mode our editorial discipline exists to refuse. Pieces that get killed leave the calendar without ceremony; the cause-of-death stays in our internal markdown for our own accountability.
Upcoming — next four weeks
-
TueApr 28
What “Context Window” Stopped Meaning
The phrase used to denote a token budget. It is now used to denote at least four different things — budget, working set, retrieval surface, and the agent’s effective horizon — and the conflation is producing bad engineering decisions. A taxonomy, with the cases each definition should be reserved for.
-
FriMay 1
The Metric That Survives Its Own Gaming
Following “Why Your Agent ROI Number Is Wrong,” the next question: which measurements degrade least under Goodhart pressure when teams are explicitly rewarded for moving them? A working answer, drawn from three operator interviews, with the limits stated.
-
TueMay 5
Tool-Return Provenance: The Field Without a Standard
Every tool an agent calls returns data the agent treats as testimony. There is no common metadata convention for marking that testimony’s source, freshness, or trust class. What partial conventions exist, where they conflict, and what an interim agent-side discipline looks like.
-
FriMay 8
Onboarding an Agent Is Onboarding a Coworker
Most teams onboard agents the way they onboard tools: credentials, scope, done. The teams getting durable performance onboard them the way they onboard people: shadowing, graduated scope, named sponsor, review cadence. Why the second pattern produces results the first does not.
-
TueMay 12
The Retrieval Layer Is the New Prompt
Prompt engineering moved from string-crafting to retrieval design somewhere in the last twelve months and most teams have not noticed. What the shift is, why prompt-engineering literature has gone quiet, and what discipline has replaced it where it has been replaced well.
-
FriMay 15
Reviewing an Agent’s Pull Request
A field manual for the human who has to approve work an agent did. Not a checklist — an account of the cognitive moves that distinguish review from rubber-stamp, with the failure modes named.
-
TueMay 19
Slot held for the week’s real disclosure
Held open for a real disclosure or technique landing in the prior two weeks. If nothing warrants treatment, falls back to a planned Context Engineering piece on eval-suite drift. Pretending to know in late April what the security landscape demands in mid-May would be calendar theater; the slot is honestly empty until it isn’t.
-
FriMay 22
What the First Quarter of Agent Coworkers Looked Like, From Three Teams
Three operators, each running agent coworkers in production for ~90 days, on what changed about their team that they did not predict. Not a vendor case study — anonymized, structured around the surprises, not the wins.
-
FriMay 29
Planned — title pending
Slot reserved from thematic work in progress. We are deliberately not pretending to know the title five weeks out.
-
TueJun 2
Planned — title pending
Slot reserved from thematic work in progress. Same discipline as the May 29 slot.
Recently published
-
SatApr 25
-
FriApr 24
-
WedApr 23
-
WedApr 23
-
WedApr 23
-
WedApr 23
Q3 2026 thematic outlook
Themes, not titles. Four candidate clusters the publication intends to work over June through August, plus a fifth slot held open for whatever the quarter teaches us we did not yet know we should be writing about. These have not been run through the four-axis editorial lens yet — they are direction-of-travel, not commitments. A lens-checked Q3 plan is owed before mid-June; clusters that cannot yield at least one piece passing the lens get dropped.
-
Q3Cluster 1
The discipline that replaced prompt engineering
What teams who used to write prompts now actually do. Retrieval design, context shaping, evaluation harness work. Sits primarily in The Substrate; the labor-organization version lives in The Operators.
-
Q3Cluster 2
Trust formation in human-agent teams over time
How does trust between a human and an agent coworker form, decay, and recover? Drawn from longitudinal interviews with operators who have run agent coworkers for two quarters or more by the time the cluster runs.
-
Q3Cluster 3
Measurement after the dashboard
A successor cluster to “Why Your Agent ROI Number Is Wrong” and “The Metric That Survives Its Own Gaming.” What measurement looks like when teams stop asking the dashboard to settle questions the dashboard was never going to settle.
-
Q3Cluster 4
The governance gap in agent-to-agent work
What happens — what is already happening — when agents call other agents and the human is two or three hops away. Treated as a community-design problem, not a policy problem.
-
Q3Reserved
Held open
Reserved for whatever the quarter teaches us we do not yet know we should be writing about. Reserving the slot is part of the editorial lens.
EDITORIAL_CALENDAR.md in the repo, including drafting/in-review states and the kill log. This page mirrors only what editorial policy commits to showing readers. Cadence: Tuesday Substrate, Friday Operators, with a floating mid-week piece every other week when an event warrants it. The Standing column appears bi-monthly.